1. Listen
VocaClear reads your microphone at 48,000 samples per second — full-band audio, the same quality professional recording gear uses.
On-device AI listens to every slice of your microphone audio, keeps your voice, and turns everything else down, before your meeting app ever receives it.
“Letter two. From John Graham, at the Union Stock Yards in Chicago, to his son, Pierrepont, at Harvard University. Mr. Pierrepont's expense account has just passed under his father's eye, and has furnished him with a text for some plain particularities.”
Real, unedited output from the VocaClear engines. Voice: “Letters from a Self-Made Merchant to His Son” by George Horace Lorimer, read by a LibriVox volunteer (public domain). Source Background noise: real field recordings by Joseph Sardin, published as CC0 (public domain) on BigSoundBank, mixed in by VocaClear. Source
VocaClear reads your microphone at 48,000 samples per second — full-band audio, the same quality professional recording gear uses.
A neural network trained on large collections of speech and real-world noise estimates, for every slice of sound, how much of it is your voice. Everything else is turned down.
Your cleaned voice is sent to a virtual microphone in real time. Zoom, Teams or any other app simply uses it as its microphone.
Traditional noise suppression learns the steady hum of a room and subtracts it. That works for a fan, but fails on the noises that matter most: a keyboard, a slammed door, a passing car or a busy café, which change from moment to moment.
VocaClear's engines are neural networks that have learned what human speech looks like across the whole frequency range. Instead of subtracting a fixed noise profile, they decide moment by moment which parts of the sound are voice. That's how they can remove sudden, unpredictable noise without making you sound robotic.
And because the networks run on your own processor, nothing is streamed to a server. Your audio stays private, works without a fast connection, and adds no network delay.
Same clips you can play above. Higher is better.
| Noise | Quieter pauses (dB) | Voice clarity (dB) | |
|---|---|---|---|
| ClarityMax | FeatherLite | ClarityMax | |
| Keyboard | 29.5 | 16.6 | +7.2 |
| Café | 21.6 | 15.5 | +4.1 |
| Air conditioner | 29.1 | 16.4 | +4.1 |
| Street | 29.1 | 16.3 | +6.7 |
“Quieter pauses” compares loudness during pauses in speech. “Voice clarity” is SI-SDR, a standard research measure that rewards removing noise and penalizes damage to the voice, for ClarityMax, the default engine. FeatherLite is the low-CPU option: it quiets pauses less and keeps a little more of the room, the trade for running on very old laptops.
Create your workspace at vocaclear.com in under a minute. You get 2 seats for 7 days, with no credit card.
In your dashboard, add a username and password for each person who will use the app. A seat is one person on one PC; someone who needs it on a second computer uses a second seat.
Install the Windows app on each PC and sign in. It also needs VB-CABLE, a separate virtual audio driver from VB-Audio: run its setup as administrator, then restart the PC.
In Zoom, Teams or Meet, choose “CABLE Output” as your microphone. That's it — everyone now hears your cleaned voice.
It gives the best quality on most computers made in the last decade.
If your PC runs hot or other apps feel slow, FeatherLite uses far less processor time.
Maximum removes the most. Gentle keeps a little room sound, which some people find more natural.
Headphones stop other people's voices from your speakers reaching your microphone.
Start your free 7-day trial with 2 seats. No credit card required.